Undoubtedly, these are tough times for the Asian foundries. Yet the past year has also been one of the most interesting, characterized by new competition, surprising alliances, dubious strategy changes and the emergence of an undeniable giant.
None of this would probably have happened without such a stark backdrop. For better or worse, the most severe downturn in semiconductor history has served up some great theater in the Far East.
First off, Taiwan Semiconductor Manufacturing Co. has shown that if the competition is relatively weak, then bad times can actually be good times. For a spin-free analysis, just follow the numbers.
In 2000, TSMC was roughly 1.5 times larger (by revenue) than its nearest rival, United Microelectronics Corp., and its profit 1.2 times higher. By the end of last year, TSMC was nearly 2.5 times larger than UMC and its profit three times higher. TSMC increased its R&D budget by about $30 million, to $338 million (excluding capital expenditures), as UMC reeled in spending by about that same amount, to $203 million.
Over the next several months, TSMC is expected to widen the gap, as it increases capacity for 0.13-micron technology and starts pilot production of 90-nanometer chips. If there was any doubt before, the company has shown it is the Intel Corp. of the foundry business-and IBM Corp.'s late arrival to the party isn't likely to change that.
While TSMC's performance has been impressive, UMC's has been the more peculiar story. It seems that by default it has become the AMD of the foundry industry, even striking up a relationship with said company because the two apparently had so much in common. Then found out they didn't.
Being second-best is not what UMC wants. But barring a catastrophic misstep at TSMC, it may be the best UMC will achieve in the foreseeable future. The company recently veered in a new direction, saying it would cull some 200 companies from its client list to better serve the remaining customers. At the same time, UMC has said it will no longer be a player on the extreme leading edge of technology, where it is bleeding money for little immediate gain.
The reaction has ranged from alarm to shock. Small companies fear being shut out; large companies question the foundry's commitment to being a technology leader, which requires being a big spender. "I don't think it's a coincidence that three to four months before they announced this [new strategy] that you had Xilinx announce an agreement with IBM and AMD run over to IBM," said Bill McClean, principal analyst at IC Insights. "You need to be on the leading edge. That's where the revenue per wafer is, and that's where the margins are. UMC has lost sight of that."
UMC says its vision is clear and that its strategy is misunderstood. Its chief technologist, Sun Shih-wei, insists that the company will be a leader in 90-nm technology and a worthy rival at the 65-nm generation. "I don't know how to change the perception," he said, "but we aren't pulling back. We are being aggressive. Our 90-nm isn't behind anybody, and we are getting excellent feedback from our customers."
As for the reduction in customer accounts, the foundry cites a condition that is shared by other top-tier foundries: Its revenue mostly comes from the top 20 customers, such as Xilinx, and not from customers 200 through 400. The end to this story will emerge over the next year or so.
The next year will also be key for three other high-profile Asian foundries: Singapore's Chartered Semiconductor Manufacturing and China's Semiconductor Manufacturing International Corp. (SMIC) and Grace Semiconductor Manufacturing Corp.
Chartered hopes to be profitable by the fourth quarter, after about two years of losses. An industry recovery would put it in the black, but it will need a successful partnership with IBM to keep it there. Without Big Blue's helping hand, most analysts believe, Chartered would not be able to compete at the leading edge.
In China, Grace will be ramping up this year, so its proving period is just beginning. SMIC, however, has been rolling wafers for more than a year and has chosen a strange path, becoming something of a memory subcontractor. With the likes of Elpida Memory Inc. and Infineon Technologies as clients, the firm is planning to finish the shell for a 300-mm wafer fab this winter and move equipment in shortly thereafter.
SMIC's predecessors have also used memory as a way to fill fabs, but not quite to the extent of SMIC. All eventually ditched memory as a utilization driver, and it's possible that SMIC may do the same, despite its early commitments. "Running commodity memory provides a way to buy nonthreatening leading-edge technology," said Len Jelinek, a semiconductor analyst with iSuppli Corp. "How can the United States argue that commodity DRAM will enhance the military capability of China?"
Jelinek also observed that while memory sales will help pay the bills early on, it's hard to sustain that forever. "I do not think SMIC wants to be a memory foundry," he said. "I believe that their current approach is a means to an end."
If a recovery does indeed arrive, then the less-talked about fabs in Asia could start to get more attention. Some say it is already happening. "We're somewhere between optimistic and wildly optimistic [for the second half]," said Steve Della Rocchetta, executive vice president of sales and marketing at Silterra Malaysia Sdn. Bhd. "Things are picking up. The question is how sustainable it is. We have been wildly optimistic for short periods in the past, but it hasn't held on."
This time feels different, though, Della Rocchetta said, adding that it reminds him of the summer of 1998. "Customers are acting like they're interested," he said. "They are concerned about capacity, and they're actively qualifying you instead of just paying lip service." He expects Silterra to turn profitable in 2004.
If things remain bad, or even mild, the small foundries will be in no-man's-land, said IC Insights' McClean. Unable to match the tech prowess of TSMC, UMC and IBM, they will also be squeezed by the double whammy of lower margins at 0.35- to 0.18-micron technology and more competition from emerging Chinese foundries, which will also play in this space.
These are, indeed, most interesting times. The Asian stage is set, and Act II (let's call it The Upturn) looks full of confrontation and intrigue.
See related chart